Issue with stack) on 6.x kernels

Issue with stack) on 6.x kernels

[Kris Van Hees]

Newer 6.x kernels (I tried 6.5.0 and 6.8.0) seem to behave differently for
stack() than e.g. 5,15,0,  Test test/unittest/aggs/tst.stack.d FAILs on those
6.x kernels whereas it passes on 5.15.0.  I am not sure yet why there is a
difference, but Eugene, perhaps you can take a look into this?

	Kris

Re: Issue with stack) on 6.x kernels

[Kris Van Hees]

On Tue, Apr 30, 2024 at 09:48:02PM -0400, Kris Van Hees wrote:
> Newer 6.x kernels (I tried 6.5.0 and 6.8.0) seem to behave differently for
> stack() than e.g. 5,15,0,  Test test/unittest/aggs/tst.stack.d FAILs on those
> 6.x kernels whereas it passes on 5.15.0.  I am not sure yet why there is a
> difference, but Eugene, perhaps you can take a look into this?
> 
> 	Kris

Here is an example:

$ sudo ./build/run-dtrace -s test/unittest/aggs/tst.stack.d 
[sudo] password for kvanhees: 
  0xffffffffc0129e86                                
  0xffffffffc0129e86                                
  0xffffffffc0129e86                                

              0xffffffffc0130250
              0xffffffffc0130250
              0xffffffffc012b2bb
              0xffffffffc012b83c
              vmlinux`schedule+0x9

The strange thing is that it actually does give good results at times.

So I am not sure what is going on.

Re: [DTrace-devel] Issue with stack) on 6.x kernels

[Alan Maguire]

On 01/05/2024 14:58, Kris Van Hees via DTrace-devel wrote:
> On Tue, Apr 30, 2024 at 09:48:02PM -0400, Kris Van Hees wrote:
>> Newer 6.x kernels (I tried 6.5.0 and 6.8.0) seem to behave differently for
>> stack() than e.g. 5,15,0,  Test test/unittest/aggs/tst.stack.d FAILs on those
>> 6.x kernels whereas it passes on 5.15.0.  I am not sure yet why there is a
>> difference, but Eugene, perhaps you can take a look into this?
>>
>> 	Kris
> 
> Here is an example:
> 
> $ sudo ./build/run-dtrace -s test/unittest/aggs/tst.stack.d 
> [sudo] password for kvanhees: 
>   0xffffffffc0129e86                                
>   0xffffffffc0129e86                                
>   0xffffffffc0129e86                                
> 
>               0xffffffffc0130250
>               0xffffffffc0130250
>               0xffffffffc012b2bb
>               0xffffffffc012b83c
>               vmlinux`schedule+0x9
> 
> The strange thing is that it actually does give good results at times.
> 
> So I am not sure what is going on.
>

here's an upstream thread discussing stacks + fentry:

https://lore.kernel.org/bpf/49b9b6f.1279.18aadb90e05.Coremail.chang-liu22@mails.tsinghua.edu.cn/

Looks like the top few entries relate to the trampoline and fentry program..

> _______________________________________________
> DTrace-devel mailing list
> DTrace-devel@oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/dtrace-devel